How Big Can Amazon Get?

by Geoff Gannon


Someone emailed me this question:

I’ve been looking at Amazon for a little short of a year and because the equity doesn't provide a good price to value ratio and margin of safety, in my view, I’ve held back from investing so far. 


There’s a little brainteaser which involves asking people how much sales, as a percentage of total sales, are done online. The numbers people answered were surprisingly high. Right now they make up 10% of total sales in the U.S. and I’m positive that number won’t stop there.


Assuming internet sales grow at 8% for the next 10 years (16% right now) one-fifth of sales will be online (also assuming 4% historical growth rate on U.S. retail sales). It would be a $1.8tn industry, x4 the size of today. Is an 8% growth rate overly optimistic given that as the base grows, percentage growth usually slows? Not looking for exact numbers, just rough measures.


Amazon has 50% market-share and I would assume that isn’t sustainable and the market’s perception of a winner-takes-all is exaggerated, many businesses could co-exist. I don’t think this is a Facebook/Google “aggregator” situation in which the viability of a new company competing with them decreases as they get more users and advertisers. The viability surely decreases to some extent but not to the extent that it does for the other two advertising companies, the retail industry is also much bigger.  


All in all, bigger but perhaps fewer companies will co-exist in the online space, and when it comes to smaller online commerce firms, they'll perhaps cater more to niches as a European pet food company is doing. However, even a 20% market-share comes out to a huge revenue number when we extrapolate what the internet sales number might be in the future. 


Very successful U.S. retailers like Walmart have ~6% market share of total retail sales, that’s significantly higher for specific categories. In terms of online though I would apply a higher than average market share for Amazon because the supply chain is significantly harder to build and optimize (e.g. more nodes) and therefore in my view this increases barriers to entry. My second question is: Do you think this is a fair assumption?


Anything I say in this post is not a suggestion to buy Amazon as a stock today. This is because of the price. We’ll get to that at the end of my article. For now, let’s just say that Amazon stock is priced high enough that unless it grows faster than a conservative estimate of how fast internet retail would grow, or it expands into other things (which, of course, it already has), or it pays out dividends, buys back stock, makes good acquisitions, etc. on top of the industry’s growth – your return in the stock could be just so-so. In other words, Amazon keeping its current position in a fast growing online retail industry isn’t going to be enough to get you great returns in the stock. This is because of the starting price of your investment.


So, that part of my answer is pessimistic.


On the other hand, I think in the very long-term you've underestimated Amazon's likely size. There are several reasons for this. 


One, I don't think judging sales nationally by category makes much sense. For example, yes, Wal-Mart may have a certain share of Sporting Goods nationwide or something. But, in reality, Wal-Mart has little presence in certain categories - for example, groceries - in some parts of the country (Northern New Jersey, Southern California, etc.) and yet has a very high share of almost all categories in places like rural Oklahoma. Supermarkets are very, very regional in the U.S. And so national market share is misleading. For example, the financial press often talks about Wal-Mart and Kroger as if they have the strongest position in groceries. In reality, companies like Publix (Florida) and HEB (Texas) probably have the strongest competitive position because they have very high market share in specific states and essentially zero presence anywhere else. In an upcoming podcast (it’ll air next Wednesday), a portfolio manager talked to me and Andrew about why he likes Village Supermarket (VLGEA). Village is a member of the Shop-Rite co-cop (called Wakefern) that has high market share in New Jersey but less presence in surrounding states and no presence in other regions of the country. However, within Village’s home state of New Jersey the Shop-Rite branded co-op has higher market share than Wal-Mart. At least as of a couple years ago, I know there were 4-5 more Shop-Rite stores in New Jersey than there were Wal-Marts that carried fresh food. Kroger has no presence in the state. And the new competition that worries New Jersey supermarket operators the most is from Wegman’s entering the state not from a national chain like Wal-Mart or Whole Foods. Wegman’s started as a New York state supermarket and expanded into adjacent states over time. It is basically in “Mid-Atlantic” states only.

This is a good indication of how misleading market share data can be. Kroger is one of the biggest supermarket companies in the country. New Jersey is the 11th most populous state in the United States, and yet Kroger has 0% market share. Wal-Mart is a leader nationally in food retail. Its market share in New Jersey is not big. There are several regional chains that all sell a lot more food in the state than Wal-Mart does. This, of course, means that Kroger and Wal-Mart have much, much higher market share in food retail in those local markets where they are the #1 or #2 chain. There are good reasons why brick-and-mortar retail breaks down so regionally. There are less good reasons why online retail would break down that way. And while Kroger and Wal-Mart are also-rans in plenty of places around the U.S. in plenty of product categories – there are quite a few cities, counties, and even states where the two chains wouldn’t have a #1 or #2 market share rank – Amazon is a leader in online retail everywhere. It’s not like Amazon is well behind other online retailers in certain states. So, the analog you should use for Amazon’s potential market share is more like Wal-Mart’s market share in general retail in the most rural counties in the U.S., or the market share in groceries of HEB in Texas, Publix in Florida, etc.


How big is that market share?


It depends on how narrowly you define the regions and the markets. If you define Publix’s market as “traditional” grocery sales (this excludes things like deep discounters) you would get a market share of 55% for the company. That 55% is over its entire trading area. It has greater than 60% market share in some local markets. But, let’s say it’s possible for a company to have 50% market share in something like groceries.


HEB operates throughout Texas. But, because Texas consists in large part of 4 metro-areas that don’t really interact that much economically – Houston, Dallas, Austin, and San Antonio are all far enough apart that they are different “regions” of the same state – a retailers market share (like a bank’s market share) will vary tremendously by which metro area you are measuring. HEB’s home market is “South Texas” this includes the Austin and San Antonio markets and goes as far North as Waco (but not Dallas). HEB has 60% market share in South Texas. Wal-Mart is second in South Texas with 27%. No other seller of groceries has a meaningful share of the market in South Texas. So, you have two companies with over 85% of the market and then you have the other 15% of the market in the hands of competitors with 2% or less market share. The best market share data I have for South Texas groceries is…HEB 60%, Wal-Mart 27%, Safeway 2%. Now, companies like Whole Foods operate profitable stores in the area. They just operate a small number of stores that have bigger market share in a very local area. Across the whole region they add up to a low market share.


But, this again points to the possibility that you are overestimating the likely fragmentation of online retail. Retail in the U.S. is often fragmented because competitors with the most successful models started in different parts of a state, country, etc. and then slowly grew to the point where they came into contact and competition with each other. For example, HEB started in South Texas and Wal-Mart started in Arkansas. HEB tried to expand into Central Texas early in its history and had problems and focused on South Texas and eventually Northern Mexico as well.


What I’m saying is that if competition in online retail is largely a local matter – you’ll get very fragmented market share nationally like you do with retailers in the U.S. But, if online retail makes all of the U.S. market more like a single local market in offline retail – then, it’s entirely possible to have market share breakdowns like: #1) 50% market share, #2) 25% market share, #3) 12% market share, #4) 6% market share…everybody else: 7% market share. And that’s very possible. It may be that the leader in online retail has 50% of U.S. online sales and the top 4 together have 90%. I can think of lots of mature local retail markets where the leader having 50% and the top 4 having 90% is not uncommon. So, it may be that when online retail is fully mature Amazon has 50% market share in the “everything store” category.


What we’re talking about here is scale. And a lot of these lists potentially mis-measure the scale that matters. For example, does having 30% market share in one town instead of 10% market share in one town really mean the same thing as going from 5% to 15% market share nationally. When buying Heinz ketchup or Budweiser beer or Kingsford charcoal it might. But, when advertising it certainly doesn’t. You can target ads locally. The fact that some people in New Jersey have heard of Kroger doesn’t help Kroger sell to people in New Jersey. So, concentration at a very local level is useful. Having good density in a state overall – moving a lot of volume of stuff that needs to go through warehouses and trucks and such – also helps. That’s one reason why you have things like the Wakefern co-op and why supermarkets always try to cluster themselves enough. In fact, the reason why some otherwise good concepts have run into trouble is often due to the failure to put enough stores close enough together in the same state. If you don’t do that, you need to rely on someone else to warehouse and move all your stuff.


Amazon isn’t at a disadvantage to anyone in delivering this to your front door. So, they have already reached the scale needed to have the ability to move stuff to you quickly and cheaply. Now, they don’t really have the last mile thing down yet. And there are certainly categories where it’s lower cost to buy in-store than online and shipped to your door. Amazon is unlikely to make as much money as PetSmart on dog food if Amazon is shipping it to Prime members and PetSmart is selling it in its stores. But, I can’t think of many categories – I can think of a few, and most are kind of unusual like very high value to weight or very low value to weight – where someone else has an advantage over Amazon in cheaply and quickly getting an item to your door (rather than to a store for pickup).


So, we have to think of scale more specifically. What specific kind of concentration are we talking about? And how does it benefit the company?


When thinking about market share, scale, etc. it helps to breakdown what kind of concentration you are thinking about. Where does this company have bargaining power? What is the actual corporate function that provides it with profit? Is it having a big share of each customer's wallet, a big share of all purchases in a certain product category, or a big share of sales in a certain region? Wal-Mart is big nationally. But, there are plenty of towns where Wal-Mart is a total non-factor. There aren't places where Amazon is a non-factor. 


What business is Amazon most similar to?


Definitely not Wal-Mart. Amazon's model is much, much closer to Costco's model. How does Costco's model differ from Wal-Mart's model?


Costco does not try to be a leading general retailer in specific towns, counties, states, the nation as a whole, etc. What Costco does is focus on getting a very big share of each customer's wallet. Costco also focuses on achieving low costs for the items it does sell by concentrating its buying power on specific products and therefore being one of the biggest volume purchasers of say "Original" flavor Eggo waffles. It sells these waffles in bulk, offers them in one flavor (Wal-Mart might offer five different flavors of that same product) and thereby gets its customer the lowest price. 


There's two functions that Costco performs where it might be creating value, gaining a competitive advantage, etc. One is supply side. Costco may get lower costs for the limited selection it offers. In some things it does. In others, it doesn't. The toughest category for Costco to compete in is in fresh food. I shop at Costco and at other supermarkets in the area. The very large format supermarkets built by companies like HEB (here in Texas) can certainly match or beat Costco, Wal-Mart, and Amazon (online and via Whole Foods stores) when it comes to quality, selection, and price for certain fresh items. But, what can Costco do that HEB can't? It can have greater product breadth (offering lots of non-food items) and it can make far, far, far more profit per customer.


Let's look at that metric.


So, a big mistake that investors make when looking at things like retailers, restaurants, movie theaters, etc. is that they think about the product being sold and never about the customer. Often, investors don't know what the customer economics are for a supermarket, Costco, Amazon, etc.


This is dumb.


It's like knowing what the net interest margin is for a bank, but not thinking about how likely a customer is to keep their deposits with the bank long-term, add to it each year, make use of all sorts of different financial services, etc. Companies know a lot about customer economics and think a lot about customer economics. Investors don't.


Amazon's big, big, big advantage - this is the key to its long-term future - is the company's customer economics. It's completely different than offline retailers. But, first let's give an example of how customer economics might work. 


Investors think about supermarkets in terms of total sales and prices in store as if people are checking the price of bananas at Wal-Mart and Kroger and deciding to buy most of their shopping list this week at Kroger but then going across the street to Wal-Mart to buy the super cheap bananas they're being offered this week. That's not how actual shopping works.


Supermarket customers in the U.S. might do something like this:


- Spend $50 per shopping trip

- Make 2 shopping trips to their primary supermarket each week


And supermarket product economics might work like this


- 25% gross margin


And customer economics might work like this


- 75% annual retention rate


These are estimates. There are stores where gross margin might be 35%, there are stores (who sell fuel, etc.) where gross margin might be 20%. There are times when customer retention rate might be 50% and other where it might be 75%.


But, this is a pretty good estimate. Now, when we think about a retailer we can think in terms of the lifetime value of a customer rather than the annual results of a store. This is a much more sensible approach. After all, when you are buying into a retail stock - you are getting millions of existing customers with your stock purchase. This is totally different than buying 200 supermarkets around the country that are opening their doors for the first time this week. There's no customer base there yet. But, if the stores are run right, in about 5 years they'll be much more valuable because they will have built up a loyal customer base. It's this loyal customer base that creates much of the value in any retail stock you're buying into as an investor.


So, $50 a visit times 2 visits equals $100 times 52 weeks equals $5,200 in annual revenue. Let's round that down to $5,000 a year. So, a loyal supermarket customer might be a household spending $5,000 a year at the local supermarket of choice. The gross profit from that household will be 25% of $5,000. That's $1,250 a year. Note that there are 12 months in a year. So, even if we round this down to $1,200 – a loyal supermarket customer is actually equivalent to a subscription business where the subscriber pays $100 a month.


Imagine that Netflix charged $100 a month for its service. That's the kind of economics you get with a loyal supermarket customer. Note this is only true up to the point where the store becomes crowded. It's not unusual for a company with 25% gross margins to have 21% to 23% SG&A costs. This leaves operating margins in the 2% to 4% range. And it may make the economics of supermarket customers very different than the economics of a Netflix subscriber. SG&A costs at a store can be high. And there are certain unavoidable costs that work like overhead cost absorption.


Basically, a low volume store ends up having a greater labor cost component in everything from cashiers to stockers per order because they need a certain base level of service regardless of volume. For reasons I discussed in my Singular Diligence write-up of Village Supermarket (VLGEA) – you can read that report and two dozen other reports like it by becoming a Focused Compounding member – a bigger store with higher inventory turns and simply greater "busyness" has all sorts of wonderful things going for it on cost, quality, etc. It's an example of a "flywheel" type business. Basically, a supermarket location that is getting more and more crowded has store economics that are spiraling upwards. It can lower total costs, increase freshness of the product, increase the store size through additions to the store. It can - over the years - lower prices, increase quality, and increase selection. Meanwhile, a supermarket location that is getting less and less crowded is in a downward death spiral. It's not just that costs can get higher. The supermarket may end up with slower moving inventory, need to cut back on employee hours, it'll start skimping on investment in the store - it won't just became economically an inferior site for the owner. It'll become less appealing to shoppers too. So, in brick and mortar retail things like return on capital for a supermarket are highly tied to store economics. You'd be surprised at how much better the economics of small, established supermarkets in great locations are versus the biggest chains in the nation. It is not true that Wal-Mart and Kroger have the best economics in the supermarket industry. There are smaller operators with better economics and there always have been. So, nationwide scale is not what drives the best returns on capital in the supermarket business. Maybe it's customer economics. Let's return to that now.


So, we said a household that's a loyal shopper at a supermarket might bring in $5,000 in revenue per year and $1,200 in gross profit. This is equivalent to $100 a month in gross profit. But, it might be as low as $8 to $16 per month per loyal customer in EBIT (this is 2% to 4% of the roughly $400 in sales per household per month). In the U.S., there are often both state and federal income taxes. For brick and mortar retailers - the tax code is pretty tough. Companies like Google aggressively avoid taxes. This isn't possible with a chain of supermarkets. You have to pay taxes to both the U.S. government and to the state governments where you have stores. The result is probably like a 25% tax rate. So, $100 in gross profits becomes like $8 to $16 in pre-tax income which becomes $6 to $12 per customer per month in after-tax profit. In reality, a MARGINAL customer can add much, much, MUCH more value to a supermarket than $6 to $12 a month after-tax. Most supermarkets are nowhere near 100% full. And they will operate below breakeven if they are pretty empty. Also, new supermarket locations will lose money at first. They might have negative cash flow for a year or two and not payback the original investment till 3-5 years down the road (and this is in the case of a successful supermarket opening).


But, let's pretend that the economics of a supermarket are this simple…


Each loyal customer adds:


$400 a month in revenue

$100 a month in gross profit

$16 a month in EBIT

$12 a month in after-tax profit / free cash flow / "owner earnings"


Basically, shareholders get:


Size of Customer Base * $12 = Monthly Free Cash Flow


Or in annual terms:


Size of Customer Base * $144 ($12 * 12 months = $144/year) = Annual Free Cash Flow 


So, we'd assume that a company will make about $150 a year per customer. A company with 10 million customers should make about $1.5 billion a year after-tax (going forward, the tax cut wasn't in effect last year so all U.S. supermarkets paid much higher taxes last year than they will this year). 


So, if a supermarket chain has a loyal base of 10 million shopping households (about 8% market share; 10 million customers / 126 million households in the U.S. equals 0.08) it should make something like $1.5 billion a year in "owner earnings" free cash flow, etc. In reality, this would be hard to check because these companies tend to use debt. And some own stores and use debt while others lease stores (instead of using debt). But, we can try to check these numbers roughly against a supermarket that has around 8% market share in the U.S. The closest company to having 8% market share of U.S. groceries is Kroger. It probably has like 7% market share or so.


So, let's check Kroger's numbers against our estimates. This is what a company with 10 million customers (8% market share) should theoretically look like based on our per customer model:


Revenue: $50 billion

Gross Profit: $12.5 billion

EBIT: $1.9 billion

After-Tax Earnings: $1.4 billion


What does Kroger really have?


Revenue: $100 billion (supermarket sales only - excludes fuel)

Gross Profit: $27 billion (does NOT exclude fuel)

EBIT: $2 billion

After-Tax Earnings: No meaningful figure (but, if taxed at 25% it would have been $1.5 billion).


In other words, our per customer economics and Kroger’s actual after-tax earnings for this year match up. Items further up the income statement don’t.


So, if Kroger had 10 million customers last year each customer would contribute


Revenue per customer: $10,000

Gross profit per customer: $1,250

EBIT per customer: $190

Earnings per customer: $150 


I've found estimates that an average family of 2 (which is almost certainly smaller than the median household size shopping at Kroger, because the median household size in the U.S. is about 2.6 people and I’d assume supermarket customers are on average bigger households than overall U.S. households because single people are the least likely to shop at a supermarket) spends around $5,000 a year on groceries. However, supermarkets - like Kroger - often get 1 to 2 times more revenue from "non-grocery" items than from grocery items. So, if a household spends $5,000 on groceries it may actually spend $10,000 a year at its favorite supermarket. So, again, our per customer model is within the realm of possibility. We might be off by 20% either way. We’re not off by 100%.


Okay. Let's just use after-tax earnings of $150 per customer as our estimate. How much is a present day customer worth to a supermarket?


Let’s look at what an “owner earnings” stream would look like for a supermarket. What is one household using this supermarket as its primary shopping destination worth to the owners of the supermarket?


This is how much the owners would make from that customer each year.


At a 50% retention rate


Year 0: $150

Year 1: $75

Year 2: $38

Year 3: $19

Year 4: $9

Year 5: $5


At a 75% retention rate


Year 0: $150

Year 1: $113

Year 2: $84

Year 3: $63

Year 4: $47

Year 5: $36


Now, your question was about Amazon. And I want to move the discussion a little closer to Amazon - and further away from supermarkets. But first we need to discuss stock market value. How much value in the stock market does a company have per customer?


A company like Kroger - so, a fairly normal U.S. public company – where the market is kind of indifferent to its future prospects will likely be valued at about 15 times after-tax profit / free cash flow etc.


So, the market value per present day customer this implies is:


Year 0: $150 * 15 multiple = $2,250


That may seem like a very strange way of looking at it. But, I want to stress just how highly the market values a supermarket customer. Basically, the "market value" of a supermarket customer is greater than $2,000. That may not be how supermarket shareholders are thinking. Cable investors often think in terms of EV/Subscriber. Supermarket investors may not. But, they’re basically paying more than $2,000 per household shopping at that supermarket when they buy into that stock.


Now, an interesting question to ask is what SHOULD determine the market value per customer. Not what does. But, what should? In other words, if we had to do a really, really long-term discounted cash flow calculation – what variables would matter most?


If two companies both have 10 million customers which company should be valued higher and why?


Two variables matter


One: Annual profit per customer

Two: Retention rate 


Basically, we're talking about a DCF here. If Company A and Company B both have 10 million customers and both make $150 per customer the company that should have a higher earnings multiple (P/E or P/FCF) should be the one with the higher retention rate.


And now we can talk Costco. 


Do you want to guess what Costco's retention rate is?


Costco's 10-K says:


"Our member renewal rate was 90% in the U.S. and Canada and 87% on a worldwide basis in 2017. The majority of members renew within six months following their renewal date. Therefore, our renewal rate is a trailing calculation that captures renewals during the period seven to eighteen months prior to the reporting date."


What this means is that if we assume a 90% renewal rate that will be an over-estimate of Costco's true retention rate. Their attrition rate is really greater than 10% a year. But, I'll simplify by assuming it's 10% a year. Now, let's look at what a company with the same exact customer economics I laid out before for supermarkets would look like if it had a customer retention rate like Costco's (90%).


Year 0: $150

Year 1: $135

Year 2: $122

Year 3: $109

Year 4: $98

Year 5: $89


Costco's earnings stream per customer looks very, very different from our theoretical supermarket example. In fact, a present day Costco customer is likely to be continuing to provide the same amount of FCF to Costco 13 years from now as a customer is likely to be providing a supermarket with a 75% retention rate in just 5 years. It's a totally different business.


Costco's customers are more loyal than your average supermarket customers. Therefore, Costco should be worth much, much more PER CUSTOMER than other stocks.


TO READ THE SECOND HALF OF THIS ARTICLE, BECOME A FOCUSED COMPOUNDING MEMBER: Use the promo code “BLOG” to take $10 a month off your membership.


To learn about our managed accounts, email Andrew at or call or text him at 469-207-5844.